Neither practice is new. Consumers encounter variable prices in airline fares, hotel rooms, ride-hailing, event tickets, gasoline, online promotions; in loyalty programs, coupons, clearance pricing, and targeted offers.
What AI changes is the speed, granularity, and data intensity of pricing decisions.
So dynamic pricing is a scaled-up version of familiar supply-and-demand pricing. Airlines raise fares as seats fill. A retailer may discount excess inventory.
Likewise, personalized pricing extends familiar couponing and loyalty discounts into potentially individualized prices. The issue is whether this is “fair.” And it is fair to note that personalized pricing raises the most issues.
Supporters argue that AI pricing can improve matching between prices, demand, and consumer circumstances, such as:
Lower prices for price-sensitive shoppers (personalized pricing)
Better inventory and availability (less waste)
More relevant offers
Convenience and responsiveness, potentially giving consumers better deals during off-peak periods or when demand is low.
Critics of personalized pricing Critics focus on opacity, exploitation and the possibility that AI turns ordinary shopping data into a tool for extracting more from each customer. Issues include:
Loss of price transparency: A listed price may no longer mean “the price everyone sees,” but rather an algorithmic estimate of what a particular shopper will accept.
Privacy and surveillance
Discrimination and proxy harms, such as systematically higher prices for vulnerable groups or neighborhoods.
Exploiting urgency or hardship (higher hotel rates when the algorithm predicts funeral or other non-discretionary travel)
Harder comparison shopping
Algorithmic coordination risk (across the board price increases by all models without any explicit human agreement, raising competition-policy issues.
Both sets of arguments are logical. There could be benefits or costs for consumers, depending on how the pricing algorithms are used.
Practice | More defensible if… | More troubling if… |
Dynamic pricing | Rules are understandable, prices reflect real scarcity or cost, and changes are not designed to exploit panic | Prices surge opportunistically during emergencies or change unpredictably to pressure quick purchases |
Personalized pricing | It delivers clear, disclosed discounts and consumers can opt out or understand the basis | It silently raises an individual’s price using inferred willingness to pay or sensitive proxies |
The important question is not simply whether prices differ. It is who benefits, who loses, and whether the practice expands access or extracts surplus.
AI makes it increasingly feasible for a business to estimate the maximum price each individual customer is willing or able to pay, and then charge accordingly.
That can be economically efficient in some circumstances, but it raises serious ethical questions about fairness, privacy and the exercise of market power.
The central issue is that personalized pricing can shift value from consumers to sellers without necessarily creating any additional value for society.
The challenge is that price discrimination is not always exploitation. In some industries, such as passenger airline travel, prices are set dynamically, based on demand.
The important question is not simply whether prices differ. It is who benefits, who loses, and whether the practice expands access or extracts surplus.
Pricing approach | Example | Ethical assessment |
Cost-based differentiation | Different delivery fees based on distance | Generally defensible when costs genuinely differ |
Volume discounts | Lower unit prices for larger purchases | Often defensible, although they may disadvantage low-income customers who cannot buy in bulk |
Group-based discounts | Student, senior, or low-income fares | Can improve access and equity |
Demand-based pricing | Higher hotel rates during a major event | Often defensible if transparent and not exploitative |
Individualized discounts | AI identifies a shopper likely to abandon a cart and offers a discount | Can benefit consumers, although it may disadvantage loyal customers who do not receive the offer |
Individualized price increases | AI predicts a shopper will pay more and raises their price | Ethically questionable, especially without disclosure or meaningful alternatives |
Vulnerability-based pricing | A model predicts desperation or urgency and charges a premium | Very ethically difficult to justify |
Discriminatory pricing | An algorithm systematically charges a protected group more without a legitimate basis
| Potentially unlawful as well as unethical |
Consider a product that “costs” $40. Assume a price-sensitive customer is willing to pay as much as $55, but is offered the product for $50. The seller makes $10 on that transaction.
Another customer (not price sensitive) is offered the product at $90, and is willing to pay up to $100. The seller's gross margin for that sale is $50.
So personalized pricing nets the seller $60 in gross profit.
Alternatively, assume the seller uses an advertised rate of $90. The first customer would not buy, while the second does. The seller makes $40 in gross profit.
If the seller charged everyone $50, both would buy, but the seller makes $20 in gross profit.
Personalized pricing might permit both customers to buy at different prices, potentially increasing total output and seller revenue. But a seller capable of extracting nearly every customer's maximum willingness to pay can capture much of the surplus itself.
The point is that greater economic efficiency does not automatically imply greater fairness, or benefit for all consumers.
Potential consumers might be harmed in other ways. Common market prices set by each seller enable comparison shopping by potential consumers.
If every customer receives a different price, several things become harder:
Comparison shopping itself
Collective bargaining by consumers
Market transparency
Competition could suffer, as competitors may have less incentive to lower prices broadly
Loss of trust if consumers see such practices as efforts to exploit their personal circumstances.
The paradox is that AI can make prices more individually efficient while making markets less transparent and potentially less competitive.
Safeguard | What it would accomplish |
Price transparency | Disclose when individualized pricing is being used, especially for material price differences |
Limits on sensitive data | Restrict using health, financial distress, precise location, or other sensitive information to set prices |
Anti-discrimination testing | Test for unjustified price disparities across protected and vulnerable groups |
Essential-service protections | Restrict exploitation of emergencies or captive customers purchasing necessities |
Meaningful competition | Make switching providers and comparing prices easier |
Consumer recourse | Allow customers to challenge erroneous data or suspicious price differences |
Auditable algorithms | Maintain records that permit regulators to examine pricing practices and their effects |
Stable public prices | Require common, accessible baseline prices in selected markets while allowing clearly defined discounts |
But transparency alone is not a sufficient protection. A consumer who needs emergency transportation, for example, may have no practical choice even if the price calculation is fully disclosed.
Likewise, banning the use of explicit sensitive attributes may not prevent discrimination if an algorithm reconstructs those attributes from other information.
Perhaps the strongest ethical objection is not that AI enables price discrimination but the fact that AI enables one-sided, largely invisible exploitation of information asymmetry.
The seller knows more about the buyer's circumstances and limits than the buyer knows about the seller's pricing strategy.
Using AI to find customers who would benefit from a lower price is easier to justify than using it to identify customers who can be made to pay more. The former arguably expands access. The latter only increases profit margins for the seller.
